A new shape descriptor 2D for Content Based Image Retrieval

Content-Based Image Retrieval (CBIR) systems have been developed to support the image retrieval based on image properties, such as color, shape and texture. In this paper, we are concerned with shape-based image retrieval. In this context, we propose a new descriptor invariant to common geometrical transformations and robust to noise to describe shapes, it is an extension of the descriptor of Berretti and al., it is based on the curvature zero-crossing points from a Gaussian smoothed boundary to obtain primitives, called tokens. We propose to describe each token by its inverse of the curvature corresponding to the radius of osculating circle, its orientation and its normalized surface. Several experiments were conducted on the widely used MPEG-7 database. The performance was measured in terms of recall and precision. The obtained results show the promising performance of our method and its superiority over descriptor proposed by Berreti and al.